Top four AI/ML use cases for service providers

July 23 2019
by Al Sadowski, Nick Patience


Artificial intelligence and machine learning technologies are no longer a glimmer on the technological horizon, or purely an academic pursuit, but rather a real-world-enabling technology with defined use cases and demonstrated ROI. Organizations are integrating AI into their products and services and imbuing machine learning into their internal workflows, with the goal of improving and automating a wide variety of business processes – from customer experience, sales and marketing to niche, vertical-specific use cases. In 451 Research's Workloads & Key Projects 2018, we quantified the top service-provider AI/ML use cases.

The 451 Take

AI/ML will play a greater role in all aspects of a service provider's business, from smarter and more efficient customer care and operations to infrastructure that is more self-healing and self-aware. It is still early, so as service providers roll out capabilities, they are diversifying their selection of AI systems and partners. Over time, this space will contract to a small set of top players that can supply the needs of most use cases.

AI/ML adoption

Currently, 35% of service providers have AI/ML-based product applications to support their customer-facing business. Within 24 months, 90% will have adopted AI and ML use cases. Telcos, systems integrators, regional cloud providers, MSPs and SaaS providers have goals to automate or optimize most customer-facing processes. The benefits that AI brings in terms of automation and optimization can be applied to a wide spectrum of operational processes for a wide variety of results. Given that a service provider is brimming with processes, tasks, analyses, procedures, and operations just waiting to be augmented and automated by machine learning, there are several ways in which the technology can bear fruit for adoptees.

By way of comparison, enterprises are AI/ML laggards, and likely waiting for standards to emerge and service providers to simplify the offerings with managed services. In 451 Research's AI & Machine Learning survey, 20% of respondents have already adopted machine learning into their organizations, and an additional 20% have the technology deployed in the proof-of-concept stage. For an additional 21% of respondents, machine learning technology is a roadmap item, with plans to implement within the next two years. Taken together, 61% of enterprises will have implemented AI/ML in some capacity over the next 24 months, compared with 90% of service providers.

Figure 1
AI/ML adoption

Use cases

MLaaS (49%) – Machine learning can be hard to understand, and considerably harder to implement for those without the necessary data science skills. Service providers that have taken on the burden are offering machine learning as a service. Offering MLaaS capabilities is an opportunity for service providers to forge a lasting enterprise relationship – the training data and models provide data gravity, but they also run the risk of competing with hyperscalers.

Capacity planning (47%) and predictive maintenance (39%) – Improving their own operations is where providers expect to see the major benefits of ML. To scale operations, service providers had turned to script-based automation rather than additional headcount for routine tasks, and now machine-learning-based automation is the next phase of smarter and more efficient customer care and predictive maintenance.

Customer service (38%) – The applications of machine learning to customer-service processes – for example, chatbots or call centers – are already advanced compared with other applications of the technology. Improved customer experiences are also an objective that closely aligns with larger digital transformation initiatives toward more personalized interactions. The cited gains in competitive advantage speak to the transformational nature of the technology – and suggest how it could soon become table stakes. Providers in India (61%) are banking on chatbots and virtual assistants – a significant number of these firms are providing customer services for companies throughout the world. Winning business based on labor cost alone is no longer sustainable.

Figure 2
Q: To which of the following use cases is your company applying machine learning? 451 Research's Q4 Advisory Report
Other use cases in the realms of cybersecurity, physical security, energy management and product recommendations are also on the radars of service providers, with varying degrees of expected return.

Buy vs. build

The market for ML applications is immature – too few vendors have written software based on service-provider requirements. As such, service providers are more interested in building than buying software. When it does come to buying, the workloads for ML/AI are varied, so service providers are using a best-of-breed approach when it comes to picking a partner. One size does not fit all; there are specialist partners for specific use cases, including telecom, retail, banking and healthcare. As for the required training data, 83% of service providers want to control the data and the training models, and that is the reason why only 16% are providing data to a third party that will build for them.

When it comes to partnering with the hyperscalers' machine learning platforms, 40% of service providers prefer Google Cloud Platform Machine Learning, followed by IBM Watson Machine Learning (36%). However, among China-headquartered service providers, IBM (44%) is the top choice, whereas the lack of interest in Google (19%) could be a hangover from Google's search engine being banned in China.

Responsibility and opportunity

The application of AI/ML technologies is starting to reshape how people work, study, travel, govern, consume and pursue leisure activities. In turn, that process will almost certainly throw up new ethical, moral, legal and regulatory challenges that have only just started to be discussed. Therefore, service providers must be prepared to explain AI systems, data security and the governance of these new technologies. As enterprises increasingly turn to service providers to manage AI/ML-based services, providers have a major opportunity to provide the potentially enormous compute resources demanded by AI and machine learning workloads, which will impact decision-making around IT infrastructure and the choice of service-provider partners. Some of the business will inevitably go to the hyperscalers, but not all, given varying concerns organizations may have about them in terms of conflicts of interest, geography and price, among other issues.